Detecting Long-Term Occupancy Changes in Californian Odonates from Natural History And

DETECTING LONG-TERM OCCUPANCY CHANGES IN CALIFORNIAN ODONATES FROM NATURAL HISTORY AND CITIZEN SCIENCE RECORDS

G. Rapacciuolo, J. E. Ball-Damerow, A. R. Zeilinger, V. H. Resh

SUPPLEMENTARY MATERIAL

TABLES

Table S1: Pairwise correlation among variables used to capture effects of seasonality, minimum temperature and total precipitation in our models.

Table S2: Final set of species modeled in this study and values for four of their biological traits

Table S3: Estimates of the effect of seasonality on probability of occupancy for 34 odonate species in California using the full dataset

Table S4: Estimates of the effect of seasonality on probability of occupancy for 24 odonate species in California using the opportunistic dataset

Table S5: Estimates of the effect of minimum temperature on probability of occupancy for 34 odonate species in California

Table S6: Estimates of the effect of total precipitation on probability of occupancy for 34 odonate species in California

Table S7: Assessment of uncertainty in models of yearly trend of species’ occupancy as a function of species’ traits

FIGURES

Figure S1: Yearly change in occupancy for habitat generalist versus specialist species

Table S1: Pairwise Spearman’s rank correlation coefficients (ρ) among Julian day, minimum temperature, and total precipitation across sets of lists analyzed in this study. Values in table indicate correlation coefficients in the full dataset, with values in brackets indicating correlation coefficients in the opportunistic dataset. Values in bold indicate statistical correlations between pairs of variables.

Table S2: Species modeled in this study and values for four of their biological traits. Temperature preference values represent mean posterior estimates of the minimum temperature parameter β2 (see eqn. 1 in the main paper) generated by our occupancy models.Values on habitat specialism, migratory behavior, and diapause were obtained from regional Odonata field guides (Manolis 2003, Paulson 2009) and supplemented by expert knowledge (D. Paulson, pers. comm.).

Table S3:Estimates of the effect of seasonality on probability of occupancy for 34 odonate species in California obtained using the full dataset of species occurrence records. For each species, seasonality effects on occupancy are estimated using the posterior mean (± 95% C.I. interval) of the linear (β5) and quadratic (β6) Julian day parameter generated using hierarchical Bayesian occupancy models (see eqn. 1 in main paper). Also presented are values of Rubin-Gelman ȓ and effective sample size (ESS) corresponding to each model parameter estimate. Bolded estimates are significant (i.e. 95% C.I. do not overlap 0).

Table S4:Estimates of the effect of seasonality on probability of occupancy for 24 odonate species in California obtained using opportunistic records only. For each species, seasonality effects on occupancy are estimated using the posterior mean (± 95% C.I. interval) of the linear (β5) and quadratic (β6) Julian day parameter generated using hierarchical Bayesian occupancy models (see eqn. 1 in main paper). Also presented are values of Rubin-Gelman ȓ and effective sample size (ESS) corresponding to each model parameter estimate. Bolded estimates are significant (i.e. 95% C.I. do not overlap 0).

Table S5:Estimates of the effect of temperature on probability of occupancy for 34 odonate species in California. For each species, temperature effects on occupancy are estimated using the posterior mean (± 95% C.I. interval) of the temperature (β2) parameter generated using hierarchical Bayesian occupancy models (see eqn. 1 in main paper). Occupancy models were run either using all odonate records (i.e. full dataset) or using opportunistic records only (i.e. opportunistic dataset). However, suitable model convergence (according to the criteria detailed in the Methods section of the main paper) could not be achieved for all species using opportunistic records only. Also presented are values of Rubin-Gelman ȓand effective sample size (ESS) corresponding to each model parameter estimate. Bolded estimates are significant (i.e. 95% C.I. do not overlap 0).

Table S6:Estimates of the effect of precipitation on probability of occupancy for 34 odonate species in California. For each species, precipitation effects on occupancy are estimated using the posterior mean (± 95% C.I. interval) of the precipitation (β3) parameter generated using hierarchical Bayesian occupancy models (see eqn. 1 in main paper). Occupancy models were run either using all odonate records (i.e. full dataset) or using opportunistic records only (i.e. opportunistic dataset). However, suitable model convergence (according to the criteria detailed in the Methods section of the main paper) could not be achieved for all species using opportunistic records only. Also presented are values of Rubin-Gelman ȓ and effective sample size (ESS) corresponding to each model parameter estimate. Bolded estimates are significant (i.e. 95% C.I. do not overlap 0).

Table S7: Assessment of uncertainty in models of yearly trend of species’ occupancy as a function of species’ temperature preference, migratory behavior, specialization, and diapause. Yearly trends were summarized using either the 2.5 or 97.5 percentile of the posterior distribution for the year parameter from each species’ model, the lower and upper 95% confidence interval bounds, respectively. Models also include family and genus as random effects. Estimated coefficients, standard error (S.E.), and t value are indicatedfor each trait.

Figure S1:Yearly change in occupancy for habitat generalist versus specialist species. Yearly changes in occupancy are the mean of the posterior distribution for the year parameter β1, as estimated from occupancy models on the full dataset. Black lines represents medians and the box shows the middle 50% yearly trend estimates for cold- and warm-adapted species.